Skip to contentSkip to conversation

ipc: fix to protect IPCS lookups using RCU

PATCH Superseded Jeongjun Park 3 participants 3 messages Feb 19, 2025

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ipc: fix to protect IPCS lookups using RCU

In shm_destroy_orphaned(), we are not performing updates to the IPCS and are only calling idr_for_each(), which can be protected by the RCU read-critical section.

And if idr_for_each() is not protected by the RCU read-critical section, then when radix_tree_node_free() is called to free the struct radix_tree_node through call_rcu(), the node will be freed immediately, and when reading the next node in radix_tree_for_each_slot(), the memory that has already been freed may be read.

Therefore, when calling idr_for_each() in shm_destroy_orphaned(), it should be modify to protect it within the RCU read critical section.

Reported-by [email protected]
Fixes b34a6b1da371 ("ipc: introduce shm_rmid_forced sysctl")
---
 ipc/shm.c | 2 ++
 1 file changed, 2 insertions(+)
ipc/shm.c +2 -0
index 99564c870084..baef5afadfb9 100644
@@ -431,8 +431,10 @@ static int shm_try_destroy_orphaned(int id, void *p, void *data)
 void shm_destroy_orphaned(struct ipc_namespace *ns)
 {
 	down_write(&shm_ids(ns).rwsem);
+	rcu_read_lock();
 	if (shm_ids(ns).in_use)
 		idr_for_each(&shm_ids(ns).ipcs_idr, &shm_try_destroy_orphaned, ns);
+	rcu_read_unlock();
 	up_write(&shm_ids(ns).rwsem);
 }
 

--

#1 lore raw

* Jeongjun Park <[email protected]> [250219 08:29]:

Thanks for looking at this.

Your subject should be v2.

Quoted 9 lines

In shm_destroy_orphaned(), we are not performing updates to the IPCS and are only calling idr_for_each(), which can be protected by the RCU read-critical section.

And if idr_for_each() is not protected by the RCU read-critical section, then when radix_tree_node_free() is called to free the struct radix_tree_node through call_rcu(), the node will be freed immediately, and when reading the next node in radix_tree_for_each_slot(), the memory that has already been freed may be read.

The way rcu and reference counting is used in shm is very odd.

Quoted 3 lines

Therefore, when calling idr_for_each() in shm_destroy_orphaned(), it should be modify to protect it within the RCU read critical section.

Did you test this with lockdep and prove rcu?

Taking the rcu lock has some extra constraints that may be violated in the call chain.

Quoted 4 lines
Reported-by [email protected]
Fixes b34a6b1da371 ("ipc: introduce shm_rmid_forced sysctl")

WBN to Cc: the person that looked at v1 for you (especially since he's already Cc'ed on the email).

Quoted 3 lines
---
 ipc/shm.c | 2 ++
 1 file changed, 2 insertions(+)

Since this is a single patch without a cover letter (which is good), you can put some stuff here that you don't want to have in the git log to be used for review, such as:

v1 link
changes since v1

Quoted 17 lines
ipc/shm.c +2 -1
index 99564c870084..baef5afadfb9 100644
@@ -431,8 +431,10 @@ static int shm_try_destroy_orphaned(int id, void *p, void *data)
 void shm_destroy_orphaned(struct ipc_namespace *ns)
 {
 	down_write(&shm_ids(ns).rwsem);
+	rcu_read_lock();
 	if (shm_ids(ns).in_use)
 		idr_for_each(&shm_ids(ns).ipcs_idr, &shm_try_destroy_orphaned, ns);
+	rcu_read_unlock();
 	up_write(&shm_ids(ns).rwsem);
 }
 
--
#2 lore raw
In reply to Jeongjun Park
Quoted 3 lines · Jeongjun Park

In shm_destroy_orphaned(), we are not performing updates to the IPCS and are only calling idr_for_each(), which can be protected by the RCU read-critical section.

Well, no, that's not true. The IPCS is updated by the callback passed to idr_for_each().

Quoted 8 lines

And if idr_for_each() is not protected by the RCU read-critical section, then when radix_tree_node_free() is called to free the struct radix_tree_node through call_rcu(), the node will be freed immediately, and when reading the next node in radix_tree_for_each_slot(), the memory that has already been freed may be read.

Therefore, when calling idr_for_each() in shm_destroy_orphaned(), it should be modify to protect it within the RCU read critical section.

This is a very complicated way of not describing what the problem is. How about:

Holding the rwsem is insufficient to protect the IDR from concurrent modification. That allows radix tree nodes to be freed while we are iterating the IDR. We can prevent this by holding the RCU read lock in addition to the rwsem.

(a really good commit message would explain why holding the rwsem is insufficient, but the way ipc uses the IDR and RCU is very complicated, and I don't remember)

Quoted 23 lines
Reported-by [email protected]
Fixes b34a6b1da371 ("ipc: introduce shm_rmid_forced sysctl")
---
 ipc/shm.c | 2 ++
 1 file changed, 2 insertions(+)
ipc/shm.c +2 -1
index 99564c870084..baef5afadfb9 100644
@@ -431,8 +431,10 @@ static int shm_try_destroy_orphaned(int id, void *p, void *data)
 void shm_destroy_orphaned(struct ipc_namespace *ns)
 {
 	down_write(&shm_ids(ns).rwsem);
+	rcu_read_lock();
 	if (shm_ids(ns).in_use)
 		idr_for_each(&shm_ids(ns).ipcs_idr, &shm_try_destroy_orphaned, ns);
+	rcu_read_unlock();
 	up_write(&shm_ids(ns).rwsem);
 }
 
--